ChatGPT vs Claude vs Gemini for Everyday Work (2026)

Picking a daily AI assistant in mid-2026 means choosing between three products that have all changed shape in the past six months. ChatGPT now shows labelled ads on its Free and $8 Go tiers — the only one of the three that does — while splitting its Pro plan into $100 and $200 tiers. Claude quietly turned its $20 Pro plan into the strongest bundle in the category: Claude Code, Cowork, Design, Science, Research and Claude for Microsoft 365 are now all included. And Gemini restructured everything at I/O 2026 — a new $100 Ultra tier, the top Ultra cut from $250 to $200, and the entry AI Plus plan cut to $4.99.

If you have been running on 2025 knowledge — "Pro is $200", "Gemini Advanced is $19.99 or $249.99, nothing between", "Claude Pro is just more messages" — every one of those is now wrong. This guide compares the three on what actually matters for daily knowledge work: what each plan costs, which model you actually get, the document/research/agent features you will use every week, and the usage caps that bite mid-afternoon.

The one-line answer: ChatGPT remains the most complete all-round assistant, Claude Pro is the best value for serious individual work, and Gemini wins if your life runs on Google Workspace or a tight budget.

Plans and what they cost

Three honest readings of that table. First, the ads: OpenAI's Free and $8 Go tiers now carry labelled ads at the bottom of answers (US pilot live since February 2026, extended to Australia, New Zealand and Canada by April); paying $8 does not remove them — only Plus and above are ad-free. Second, at the $20 mark the bundles diverge sharply: ChatGPT Plus buys you deep research, agent mode and bigger caps; Claude Pro buys a whole suite of work products on top of higher limits; Google's $19.99 AI Pro throws in things that are not AI at all — 5 TB of storage, YouTube Premium Lite and $10/month of Cloud credits. Third, all three now converge on identical $100/$200 power tiers — a pricing structure that simply did not exist a year ago.

For readers in Singapore: Google's subscriptions page rendered live SGD prices to us of S$6.98 (AI Plus), S$28.99 (AI Pro) and S$139.99/S$289.98 (Ultra). USD figures for Gemini are per Google's official announcements.

ChatGPT: deepest feature set, new caveats

The everyday default is GPT-5.5 Instant, which auto-routes complex prompts to a Thinking mode; the new flagship GPT-5.6 family (Sol, Terra, Luna) announced on 9 July 2026 is still rolling out, with Sol — the strongest, aimed at coding, science and agentic work — offering manual Medium/High/Extra High reasoning on eligible plans. The catch: Sol is not available to Free or Go users at all.

For daily work the toolkit is the broadest of the three: Canvas for side-by-side writing and code editing, Projects with project-scoped memory, deep research and agent mode from Plus up, and the new ChatGPT Work agent that produces documents, spreadsheets, slides and Sites — in the desktop app, Chat, Work and Codex are available on every plan including Free. "Apps" (connectors, renamed in December 2025) hook in Google Drive, Gmail, Outlook, Teams, SharePoint, GitHub and more — but are not available in the EU.

The limits bite in specific ways. Free users get a 16K context window for the default model and 3 file uploads per day. Go and Plus get up to 160 GPT-5.5 Instant messages every 3 hours — then chats silently drop to GPT-5.5 Instant mini until the window resets. Go's Thinking mode is capped at 10 messages every 5 hours. Contexts stay small on cheap tiers: 32K on Plus for the default model, versus 128K–400K on Pro.

The honest weakness: ChatGPT is the only assistant here showing ads, and the small default contexts on Free and Plus look dated next to Gemini AI Pro's 1M and Claude's paid 200K–1M windows.

Claude: the strongest paid bundle

The default model on Free and Pro is Claude Sonnet 5 (June 2026); the most capable widely released model is Claude Fable 5, globally available since 1 July 2026 after US export controls were lifted, with Opus 4.8 in the lineup. Extended thinking is listed even on the Free plan. Anthropic does not publish a precise model-to-plan map beyond "Pro adds additional models", which is worth knowing before you subscribe for a specific model.

The work story is where Claude has moved furthest. File creation, code execution and Artifacts work on all plans including Free; memory from chat history is free too. Pro at $20/month ($17 annual) bundles Claude Code, the Cowork desktop agent for multi-step tasks, Design, Science, unlimited Projects, Research (its agentic multi-search with citations) and Claude for Microsoft 365 — Excel, PowerPoint and Word add-ins are GA on all paid plans, Outlook is in beta, and a Microsoft 365 connector (search across SharePoint, OneDrive, Outlook, Teams) works on every plan including Free. Voice mode is in beta on all plans. Slack users should note the official app becomes "Claude Tag" on 3 August 2026.

The limits are the most opaque of the three: usage resets every 5 hours, with weekly caps on top — Max plans carry two separate weekly caps — and Anthropic publishes no absolute message numbers, so budgeting heavy use means learning by hitting the wall. Paid contexts are generous: up to 1M tokens on the newest models, 500K or 200K on others.

The honest weakness: Claude has no native image or video generation on any plan, and Research — its deep-research equivalent — is paid-only while Gemini offers Deep Research even on Free. It is a pure assistant subscription: no storage, no media perks.

Gemini: the ecosystem play

Free users get Gemini 3.5 Flash as the default with varying access to 3.1 Pro; paid plans get Gemini 3.1 Pro outright, and the Deep Think reasoning tier plus the 24/7 Gemini Spark agent land first on Ultra — US only and English only for now. Gemini Omni, the new multimodal creation model, arrived at I/O 2026.

The free tier is the most generous of the three on features: Deep Research, Canvas, Gemini Live voice, Gems (custom assistants) and NotebookLM are all included at $0. Paid plans buy integration rather than new toys — Gemini inside Gmail, Docs and Vids on AI Pro, the AI Inbox rolling out to Plus and Pro, a Daily Brief agent, plus the non-AI sweeteners: storage from 400 GB (Plus) to 20 TB+ (Ultra), YouTube Premium, Cloud credits, Google Home Premium. If your organisation lives in Google Workspace, no other assistant is woven this deeply into the tools you already have open.

The limits are honest about only one thing: context windows — exactly 32K tokens on Free, 128K on AI Plus, 1M on AI Pro and Ultra. Everything else is published as multipliers (2x on Plus and 4x on Pro over free-tier limits, and "5X or 20X higher" than the Pro plan for Ultra) with no absolute numbers anywhere. Heavy features like media generation and Deep Research consume allowance faster, and free-tier Deep Research "may be unavailable during periods of high demand."

The honest weakness: you cannot know your real message budget on any Gemini plan, the 32K free context is the smallest headline paid-ecosystem entry point here besides ChatGPT Free's 16K, and the flagship features — Deep Think, Spark — are geofenced to the US in English.

Which one for which job

Two cross-cutting notes. Microsoft-centric teams have a surprising best option in Claude — its Microsoft 365 connector works even on the free plan, and Anthropic's models are now selectable inside Microsoft 365 Copilot itself; there is no official consumer "ChatGPT in Office" product. And EU-based readers should discount ChatGPT's connector story entirely until apps ship there. If subscriptions aren't your thing, our self-host roundup covers the local-model route.

What this comparison doesn't cover

This is a consumer/individual-plan comparison. We have not scored enterprise and team tiers (ChatGPT Business at $25/user/month exists; all three sell org plans with different admin, security and data-handling terms), API pricing (a completely separate cost model), or regional availability quirks beyond the ones flagged above — and model lineups are churning fast enough in 2026 that the "default model" rows have a short shelf life; GPT-5.6 was announced days before we went to press.

All claims were verified in July 2026 against official pricing and help pages. Anthropic and Google pages were fetched and read directly; openai.com and help.openai.com block automated fetching, so OpenAI figures were verified via official-page snippets — see the linked sources. Gemini USD prices are per Google's official June 2026 announcements (its live subscriptions page geo-rendered SGD prices to us, which we have listed for Singapore readers). Anything we could not confirm on an official page — including exact per-plan deep-research allowances — was left out.